Articles — Places and travel

Places and travel

AI place answers come from different directories

ChatGPT, Perplexity, and Gemini can each answer a place question with different recommendations, because each one looks up places in its own directory.

You asked two assistants for a quiet place near the station. One named three tourist favorites. The other named a different trio, with Yelp links attached.

Your regular spot never appeared in either list.

There is no single local index

Product answers often flow through shopping feeds and merchant data. Place answers flow through directory partnerships and search indexes that differ by company.

Last checked: October 2026. Partnerships change. Treat the table below as a working picture, not a verified map inside each company.

Assistant (typical) Local place data often comes from
ChatGPT (search-backed) Foursquare Places API for many location results (Foursquare, December 2024); Yelp reviews, ratings, and photos (Axios, July 2026); table booking via Yelp Reservations in the US and Canada (Yelp)
Perplexity Yelp for many restaurant-style answers; Tripadvisor for hotels and travel inventory (The Verge, January 2025)
Gemini Google Maps and Business Profile grounding in consumer and API products, which is not the same path as ChatGPT’s Yelp and Foursquare stack
Copilot Bing-shaped results; citations often mix restaurant sites, Yelp, and other directories rather than one shared list

OpenAI’s ChatGPT search announcement says search uses third-party providers and partner content without listing every partner on one page. Confirm the Foursquare and Tripadvisor links above if a partnership claim is load-bearing for your own listing. The Yelp row is the public description vendors and OpenAI have given. It is not a percentage from an SEO blog.

The directory is the product page

For a shop, the product page, the JSON-LD block (structured facts machines can read), and the product feed must agree. For a place, the record in the directory that chatbot actually used is the page that tool can recommend from.

Optimizing only a Google Business Profile helps Gemini and Google-shaped surfaces. It does not automatically move ChatGPT, which may never treat Maps as its primary local source. In the US, Yelp weight is real for ChatGPT and Perplexity; in many European cities Yelp coverage is thinner, and fame plus web guides still dominate.

That regional gap matters. Advice written for US restaurateurs is not a universal recipe. If you are in Europe and Yelp is thin in your city, put the same name, address, and hours on Google Business Profile, Apple, and your own site, and treat Yelp as optional rather than the main listing.

Famous places still win inside each directory

Even within one directory, corroboration and review volume skew toward places that already dominate. Licensed data does not erase the tourist-list effect from “best things to do near me”. It only changes which famous places each tool can retrieve.

If you run a place

  1. Know which directories your category uses in your country — Google, Apple, Bing, Foursquare, Yelp, Tripadvisor — and keep name, address, and hours aligned.
  2. Say clearly on those listings what you are for, so constrained asks can match.
  3. Do not assume one platform’s update fixes every assistant.

The listing still said open

Illustrative, not a logged test. The assistant said the restaurant was open. You arrived to a dark room and a paper sign. Often the system retrieved a listing on Maps, Yelp, Foursquare, or a tourism page whose hours were wrong, and treated that record as current enough to recommend. This is the places version of wrong price, wrong shop, fake discount.

You can win on reviews and still lose the visit if open hours, a holiday closure, or the cuisine tag is stale. The same miss can leave you named and still not booked when the assistant never had a clear next step. See you can be recommended and still not get chosen.

Stale place data hides in a few records:

When the plan depends on tonight, treat hours in the chat as a hint. Confirm on Maps, the venue site, or a phone call. That is the same habit Why AI won’t just say it doesn’t know describes: check a sure-sounding fact before you act on it.

If you run the venue, mark a temporary closure on every listing you appear in, not only on the platform you personally check. Structured LocalBusiness markup on your site can help crawlers that parse it. Use Restaurant when that is what you are. Name, address, and hours in that block should match your Google Business Profile and your other listings. It does not replace directory records ChatGPT or Perplexity license directly. Write the hours you mean on one dated note before you compare the listings, the same habit as in Wrong price, wrong shop, fake discount.

Read next if: Why you got the same five landmarks · Find out why you were skipped before you rewrite the page